Text Analysis of Jurassic Park

Sam Sitting on Jurassic Park Logo, Photoshop Courtesy of Sam Lance

1 Data Description and Citation

The data used in this analysis is the text of the book “Jurassic Park” by Michael Crichton. The data was obtained from the website Readers Library (https://readerslibrary.org/wp-content/uploads/Jurassic-Park.pdf). The citation for the book is as follows:

Crichton, Michael. Jurassic Park. New York: Alfred A. Knopf, 1990.

2 Analysis

This analysis will focus on the text of the book Jurassic Park by Michael Crichton. The book begins by following the travels of a group of scientists to Isla Nublar, and over time shifts into an action novel/ mediation on the ethics of science. I noticed this stark change in tone when reading the book, which made me curious about whether this trend would appear through text and sentiment analysis.

One quick note on the structure of this analysis is that while the book has traditional chapters, it is also broken up into six different fractal iterations. Each iteration shows a picture of a fractal, a repeating geometric pattern, and a description of how patterns begin to emerge as more of the pattern is seen. These special markers appear at pivotal moments of the book, and will be used in place of chapters to measure how sentiments change further through the book.

Example Fractal Image from the First Chapter of the Book, Courtesy of @TheRealDoctorT on X

3 Steps of Analysis

The steps taken in this analysis are as follows:

1. Load Data
2. Wrangle Data
   1. Load in PDF
   2. Convert into Dataframe
   3. Convert Iterations into Chapters - iterations best way to divide not actual chapters
   4. Separate Out Each Word
   5. Remove Other Words - names, numbers, etc.
3. Create Column Chart
4. Create Word Cloud

library(tidyverse)
library(ggplot2)
library(dplyr)
library(here)
library(janitor)
library(tsibble)
library(feasts)
library(fable)
library(tidytext)
library(pdftools)
library(ggwordcloud)
library(textdata)

4 Loading PDF Document

jp_text <- pdftools::pdf_text(here("posts","2025-03-19-jp-244", "data", "jurassic_park.pdf"))

5 Wrangling and Turning into Dataframe

#######################################################
#CONVERT INTO A DATAFRAME 
#######################################################

jp_lines <- data.frame(jp_text) |>
  mutate(page= 1:n()) |> #create a column for each page 
  mutate(text_full = str_split(jp_text, pattern= '\\n')) |> #new vector within each column, each /n keeps as its own unique element within the vector itself, need // is telling r to look for \n exactly as written
  unnest(text_full) |> #extract each element of the vector as a row in the dataframe, unique observation for each line 
  mutate(text_full = str_trim(text_full))  #remove any leading or trailing white space

#######################################################
#GET CHAPTER NAMES
#######################################################

jp_chapts <- jp_lines |> 
  slice(-(1:443)) |> #gets rid of preface + junk, find line by looking at the actual data 
  mutate(iteration = ifelse(str_detect(text_full, regex("Iteration", ignore_case = FALSE)), text_full, NA)) |> #find where says chapter, if it does, put the chapter number in the new column, if not, put NA
  fill(iteration, .direction = 'down') |> #fill in the nas with chapter from above it 
  separate(col = iteration, into = c("num", "it"), sep = " ")  |>#seperate out the word chapter and the number 
  mutate(num_numeric = case_when(
    num == "First" ~ 1,
    num == "Second" ~ 2,
    num == "Third" ~ 3,
    num == "Fourth" ~ 4,
    num == "Fifth" ~ 5,
    num == "Sixth" ~ 6,
    num == "Seventh" ~ 7,
    TRUE ~ NA_real_  # Any other values that don't match will be assigned NA
  )) |>
  drop_na()

#######################################################
#SEPERATE OUT EACH WORD
#######################################################
jp_words <- jp_chapts |> 
  unnest_tokens(word, text_full) |> #take info from text_full and put into column called word
  select(-jp_text) #taking out the column of the actual hobbit text 

#######################################################
#GET RID OF STOP WORDS + NUMBERS + NAMES 
#######################################################
jp_wordcount <- jp_words |> 
  count(num_numeric, word) #count the number of times each word appears in each iteration

jp_words_clean <- jp_wordcount |> 
  anti_join(stop_words, by = 'word') |>#take out all stop words by the column word 
  filter(!str_detect(word, regex("[0-9]", ignore_case = FALSE))) |> #take out all numbers
  filter(!(word %in% c("grant", "gennaro", "hammond", "tim", "lex", "wu", "muldoon", "malcolm", "arnold", "regis", "harding", "nedry", "ed", "dodgson", "john", "alan", "ellie", "murphy", "satler", "dr.", "henry", "robert", "bob", "morris", "manuel", "dr", "bowman", "tina", "guitierrez", "ellen", "mike", "it’s", "don’t", "that’s"))) #take out all names

Creating Column Graph of Top Words

top_5 <- jp_words_clean %>% 
  group_by(num_numeric) %>% 
  arrange(-n) %>% 
  slice(1:5) %>%
  ungroup() 

# Make the plot with reordered bars within each facet
jp_column <- top_5 %>%
  ggplot(aes(y = word, x = n)) +  # Reorder 'word' by 'n' on the y-axis
  geom_col(fill = "#520101") +
  theme_minimal() + 
  labs(title = "Top 5 Words in Each Iteration", x = "Word Count", y = NULL) +
  facet_wrap(~num_numeric, scales = "free") +  # Facet wrap by num_numeric
  theme(axis.text.x = element_text(hjust = 1))  # Optional: Rotate x labels

# Display the plot
jp_column

• The words used closely mirror the major problems faced by the characters throughout the book, especially the transition between tyrannosaur to raptor to nest

• If character names were not eliminated from the analysis, the top 5 words for every category would just be character names

Wordcloud

top_100 <- jp_words_clean %>% 
  arrange(-n) %>% 
  slice(1:100)

cloud <- ggplot(data = top_100, aes(label = word)) +
  geom_text_wordcloud(aes(color = n, size = n), shape = "diamond") +
  scale_size_area(max_size = 6) +
  scale_color_gradientn(colors = c("#520101","#a60000","#ff0000")) +
  theme_minimal()
cloud

• Similar results to the column graph, but shows how frequently the word animals is used throughout the text, with dinosaurs as a close second

• The word head also appears frequently throughout the book, likely referring to animals moving their heads to look at the characters or peeking around corners

Sentiment Analysis

#LOAD IN + LINK THE BING LEXICON
bing_lex <- get_sentiments(lexicon = "bing")

jp_bing <- jp_words_clean %>% 
  inner_join(bing_lex, by = 'word') 

#CREATE LOG RATIO FOR THE ENTIRE BOOK
bing_log_ratio_book <- jp_bing %>% 
  summarize(n_pos = sum(sentiment == 'positive'),
            n_neg = sum(sentiment == 'negative'),
            log_ratio = log(n_pos / n_neg))

#CREATE THE LOG RATIO FOR EACH CHAPTER 
bing_log_ratio_ch <- jp_bing %>% 
  group_by(num_numeric) %>% 
  summarize(n_pos = sum(sentiment == 'positive'),
            n_neg = sum(sentiment == 'negative'),
            log_ratio = log(n_pos / n_neg)) %>%
  mutate(log_ratio_adjust = log_ratio - bing_log_ratio_book$log_ratio) %>%
  mutate(pos_neg = ifelse(log_ratio_adjust > 0, 'pos', 'neg'))

#CREATE PLOT 
ggplot(data = bing_log_ratio_ch, 
       aes(x = log_ratio_adjust,
           y = fct_rev(factor(num_numeric)),
           fill = pos_neg)) +
           # y = fct_rev(as.factor(chapter)))) +
  geom_col() +
  labs(x = 'Adjusted Log (Positive/Negative)',
       y = 'Iteration (Section) Number',
       title = "Sentiment Analyis of Iterations in Jurassic Park") +
  scale_fill_manual(values = c('pos' = 'darkgreen', 'neg' = 'darkred')) +
  theme_minimal() +
  theme(legend.position = 'none')

• As was expected, the second half of the book is significantly more negative then the first half

• Looking at the text analysis, the tyrannosaur appears for the first time in iteration 4, which could be the driving factor in the tone of the book shifting